The Application Research For Network Fault Diagnosis Based On Rough Set And Clonal Sel’ection Algorithm

Computer network has penetrated into every corner of life and work, which is the basis of the information society. However, existing network fault diagnosis system is weak in real-time diagnostics and dynamic adaptability, deficient in self-learning ability and so on. Therefore, it is imminent to develop more intelligent fault diagnosis system. This paper presents the method of rough sets and clonal selection algorithm combined to solve problems above.On the one hand, the use of rough set reduction features removes the redundant attributes of network failures in order to reduce the training and clonal selection algorithm time of diagnosis; On the other hand, the use of attribute importance makes the accuracy of diagnosis higher when troubleshooting. This paper mainly involves the training phase fault detector and fault diagnosis stage:(1) The research is proposed to speed up fault diagnosis rate, improve the diagnostic accuracy of the method. First, attribute reduction of the network fault samples using rough set to extract the fault diagnosis decision-making rules is to avoid a large interference of unrelated attributes. This step can reduce the dimension of attribute failure to achieve the purpose of streamlining. Meanwhile, in the fault diagnosis period, it needs to determine various attributes of the fault samples need to collect. However, the importance degree of each attribute is not the same and there are different roles. If attribute determination with rough set is objective, the more accurate diagnosis will be, which will avoid strongly subjective drawbacks brought by attribute importance given to human.(2) The research is proposed to train each failure mode detector with the improved clonal selection algorithm and use the detector to diagnose network faults. Most existing clonal selection algorithm convergence rate is relatively slow, the detector generated is blind and lack of dynamic adaptability, resulting in a more serious misdiagnosis phenomenon. In this paper, inadequacies will be improved, for example, increasing the detector dynamic update strategy for timely insight into the transformation between autologous and non-autologous; increase the gene pool, in order to improve the quality and generation rate of the initial detector; When a new fault occurs, you can re-training the corresponding failure mode detector, in order to enhance the ability to learn; through memory and mature dual detector detection, so that more accurate fault classification.View of the rough set and the advantages of clonal selection algorithm, in particular to their effective combination. Meanwhile, mentioned for the examination of this subject to improve the effectiveness of the proposed method, simulation results in this analysis, and validation results show that the improved method proposed in this subject than the existing methods, the relative strength of its effectiveness, the diagnosis can be solved to some extent the problem of bottlenecks in the system to improve application performance.